Pixel Value Difference based Face Recognition for Mitigation of Secret Message Detection

Abstract

Data security is an important aspect of the modern digital world. Authentication is necessary for the prevention of data from intruders and hackers. Most of the existing system uses textual password which can provide only single-layer security. The textual passwords are simple but they may prone to spyware as well as dictionary attacks. Hence there is a need for a highly secure and multilayer security method. Steganography, the art of hiding the existence of a message by embedding it into another medium, can be exploited in an authentication system. Steganography has emerged as a technology that introduced steganalysis to detect hidden information. In this approach, the multimedia file is the input that is to be transferred over the media. On the transmitter side, the audio and video files are extracted. The secret audio file is embedded with an audio file using the LSB method while the face of the authenticated person is embedded into the video frame using the Pixel Value Differencing (PVD) method. At the receiver side, the face is extracted using the reverse PVD method and authenticated using the Convolutional Neural Network-based face recognition method. After authentication, the secret audio is extracted using the reverse LSB method. The results show that the MSE, RMSE, PSNR, and SSIM of 0.0000045303, 0.0021, 53.5877, and 0.9957, respectively.